Olmo-3.1-32B-Instruct-DPO: benchmark results
Allen AI's Olmo 3.1 32B Instruct DPO, pretrained on the Dolma 3 dataset and post-trained on Dolci data, DPO-aligned from its own SFT checkpoint in the Olmo 3 Instruct line. Provider: Allen AI. Released 2025-12-23. Access: Open.
Aggregate Rating 1481 ± 23, rank #845 of 1656 rated models, from 332 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EuroEval Catalan NLU - MultiWikiQA CA | 75.98 | Character F1 (%) | 93.8 |
| EuroEval Dutch Summarization - Wiki Lingua NL | 35.07 | chrF3++ (%) | 92 |
| EuroEval Danish NLU - Multi Wiki QA DA | 79.52 | Character F1 (%) | 91.8 |
| EuroEval Romanian NLU - MultiWikiQA RO | 73.9 | Character F1 (%) | 91.7 |
| EuroEval English NLU - ScaLA EN | 61.04 | Matthews correlation coefficient (%) | 91.1 |
| EuroEval Swedish NLU - Multi Wiki QA SV | 78.05 | Character F1 (%) | 90.5 |
| EuroEval Portuguese NLU - ScaLA PT | 40.23 | Matthews correlation coefficient (%) | 89.8 |
| EuroEval Portuguese NLU | 60.24 | NLU Average Score (%) | 88.9 |
| EuroEval Ukrainian NLU - Cross Domain UK Reviews | 61.48 | Matthews correlation coefficient (%) | 88.8 |
| EuroEval French NLU - ScaLA FR | 53.13 | Matthews correlation coefficient (%) | 88.6 |
| EuroEval Italian NLU - ScaLA IT | 43.24 | Matthews correlation coefficient (%) | 88.5 |
| EuroEval French NLU - Allocine | 95.22 | Matthews correlation coefficient (%) | 88.4 |
Interactive version: theaggregate.ai/model?slug=olmo-3-1-32b-instruct-dpo · How It Works · Data refreshed daily, snapshot 2026-10-20.